Daniel (Weixiang) Wan

AI Support · Agent Evaluation

Engineering-support answers need to be checked against expert references.

Fluent engineering-support answers can still miss important facts, so I connected evaluation that compares eligible Teams support answers with their linked expert references. Fixed golden cases run separately in offline CI to check whether changes reintroduce known problems.

Read case study →
Teams answerEligible linkExpert answerCompareOnline scoreFit · part · gapGolden casesOffline CISeparate path